Loading...
Please wait, while we are loading the content...
Similar Documents
Approximate Nearest Neighbor Field Computation via k-d Trees
| Content Provider | Semantic Scholar |
|---|---|
| Copyright Year | 2016 |
| Abstract | An Approximate Nearest Neighbor Field (ANNF) describes the coherency between two images A and B by approximating the nearest neighbor from image B for every pixel patch in image A. In this thesis we propose an algorithm using k-d trees and PCA to efficiently compute an ANNF between two images. This approach is then compared to a state-of-the-art method called PatchMatch which tackles this problem in a different way. Because both methods exploit different aspects of the data, it is not directly clear which method is more suited for ANNF computation. This research aims to provide a better insight in this area. What we find is that PatchMatch yields reasonable accuracy about 3-4 times as fast as our approach, but when given enough time a k-d tree + PCA will surpass accuracy of PatchMatch. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://www.cs.ru.nl/bachelorscripties/2016/Jeftha_Spunda___4174615___Approximate_Nearest_Neighbor_Field_Computation_via_k-d_Trees.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
